With the popularization of artificial intelligence technology, face recognition systems have been widely used in various fields. However, face recognition systems still face evolving challenges, particularly vulnerabilities to presentation attacks such as print and electronic screen display attacks. To address these issues, in this paper, we proposes a sophisticated multi-view anomaly detection approach for face anti-spoofing. Unlike conventional binary classification methods, our approach employs a perspective-based reconstruction model leveraging multi-view parallel networks. This framework incorporates spatial and textural information, enhancing the model’s capability to capture diverse data distributions associated with various attack modalities. Rigorous evaluations on NUAA, MultiSpectral-Spoof, and LCC-FASD datasets validate our approach, showcasing substantial improvements over prevailing state-of-the-art methods. These results underscore the adaptability and efficacy of our proposed model in addressing contemporary face recognition challenges.